week ending

PRL 102, 236805 (2009) PHYSICAL REVIEW LETTERS 12 JUNE 2009

Defect Scattering in Graphene

Jian-Hao Chen,1,2 W. G. Cullen,2 C. Jang,2 M. S. Fuhrer,1,2 and E. D. Williams1,2
1
Materials Research Science and Engineering Center, Department of Physics, University of Maryland,
College Park, Maryland 20742 USA
2
Center for Nanophysics and Advanced Materials, Department of Physics, University of Maryland,
College Park, Maryland 20742 USA
(Received 23 March 2009; published 12 June 2009)
Irradiation of graphene on SiO2 by 500 eV Ne and He ions creates defects that cause intervalley
scattering as is evident from a significant Raman D band intensity. The defect scattering gives a
conductivity proportional to charge carrier density, with mobility decreasing as the inverse of the ion
dose. The mobility decrease is 4 times larger than for a similar concentration of singly charged impurities.
The minimum conductivity decreases proportional to the mobility to values lower than 4e2 =
h, the
minimum theoretical value for graphene free of intervalley scattering. Defected graphene shows a
diverging resistivity at low temperature, indicating insulating behavior. The results are best explained
by ion-induced formation of lattice defects that result in midgap states.

DOI: 10.1103/PhysRevLett.102.236805 PACS numbers: 73.63.
b, 72.10.Fk, 81.05.Uw

The strong carbon-carbon sp2 bonds which provide 2e2 n 1

graphene with high intrinsic strength [1] and make possible c ¼ nec ¼ ; (1)
h nc Gð2rs Þ
the isolation of single atomic layers [2] also result in a very
low density of lattice defects in graphene prepared by where e is the electronic charge, h the Planck’s constant, nc
mechanical exfoliation [3,4]. However, lattice defects in the charged-impurity density, rs the Wigner-Seitz radius,
graphene are of great theoretical interest [5,6] as a potential and Gð2rs Þ an analytical function of the dimensionless
source of intervalley scattering, which in principle trans- interaction strength in graphene. For graphene on SiO2 ,
forms graphene from a metal to an insulator [7,8]. Lattice Eq. (1) gives c  5  1015 V
1 s
1 =nc [19,20]. The ran-
defects are also likely to be present in various concentra- dom charged-impurity potential also gives rise to electron-
tions in graphene synthesized by reduction of graphene hole puddles with a characteristic intrinsic carrier density
oxide [9,10], chemical vapor deposition [11,12], or segre- n , which is a function only of nc , d (the impurity-graphene
gation of carbon on the surface of SiC [13]; hence, it is distance,) and rs , resulting in a minimum conductivity
important to understand their impact on electronic min ¼ n ec . However, strong disorder, modeled as a
transport. deep potential well of radius R, is predicted to produce
Here we show that ion-irradiation-induced defects in midgap states in graphene [6], and a conductivity which is
graphene cause a significant intensity in the Raman D also roughly linear in n [5]:
band associated with intervalley electron scattering [14–
16] and give rise to a constant mobility, similar to the 2e2 n 2
effect of charged impurities, but with a magnitude 4 times d ¼ ned ¼ ln ðkF RÞ; (2)

h nd
lower than for a similar concentration of singly charged
impurities. This result is in contrast to the carrier-density- where nd is the defect density and kF is the Fermi wave
independent conductivity for weak point disorder [17,18] vector. A third type of scattering in graphene, weak point
but consistent with the theory of strong scattering by disorder, is predicted to give rise to a carrier-density-
midgap states [5,6]. Unlike charged impurities [19], la- independent resistivity s [17], which has been observed
ttice defects (1) do not change the residual charge den- experimentally [18].
sity in electron-hole puddles, (2) greatly depress the To investigate the dependence of graphene’s conductiv-
minimum conductivity, even below 4e2 =
h (the theo- ity on defect density, cleaned graphene on SiO2 was irra-
retical minimum value of the conductivity at the Dirac diated with 500 eV Heþ and Neþ in ultrahigh vacuum
point in the absence of intervalley scattering [7]), and (UHV) at low temperature (10 K for Heþ irradiation and
(3) induce insulating temperature dependence of the 40–80 K for Neþ irradiation), and the conductivity mea-
conductivity. sured in situ in UHV to prevent subsequent reaction with
Transport with constant mobility is predicted for both molecules from the gas phase [21]. Ion irradiation of
charged-impurity scattering and scattering by midgap graphite at these energies produces one atomic-scale de-
states. Charged-impurity disorder in graphene results in a fect, most likely a carbon vacancy possibly with a trapped
conductivity noble-gas atom, per incident ion [22,23].

0031-9007=09=102(23)=236805(4) 236805-1 Ó 2009 The American Physical Society

 week ending
PRL 102, 236805 (2009) PHYSICAL REVIEW LETTERS 12 JUNE 2009

a) Pristine Graphene
Raman Intensity [a.u.] G'
G

D
+
b) G After Ne Irradiation
G'

1200 1350 1500 1650 2400 2550 2700 2850

-1
Wave Number [cm ]

FIG. 1 (color online). Raman spectra (wavelength 633 nm) for FIG. 2 (color online). Conductivity vs gate voltage curves for
(a) pristine graphene and (b) graphene irradiated by 500 eV Neþ pristine graphene and following Neþ ion irradiation doses with
ions at a dose of 1012 cm
2 . cumulative exposures indicated. Irradiation and measurements
were performed at T ¼ 41 K in a ultrahigh vacuum. The dashed
curve shows the predictions from Eq. (2) with the experimentally
Figure 1 shows the Raman spectra, taken under ambient
at defect density nd ¼ 7:22 
extracted defect radius R ¼ 2:3 A
conditions, for a representative graphene sample before 1011 cm2 .
irradiation, and after irradiation by Neþ at a dose of
1012 cm
2 (1 Neþ per 4  103 carbon atoms). The pris-
tine sample shows a Lorentzian G0 band characteristic of runs and 9:3  10
16 V s for the Heþ irradiation runs and
single layer graphene, and no detectable D band. Upon an offset that yields the initial mobility in the graphene
irradiation, the appearance of the D band indicates sig- prior to each irradiation run [21]. Assuming midgap scat-
nificant intervalley scattering [14,16]. A very rough esti- tering [Eq. (2)], at carrier density n ¼ 2  1012 cm
2 , the
mate of the defect spacing can be made using the empiri- proportionality constant yields the defect radius R ¼ 2:3 A

cal formula La ¼ ½2:4  10
10 nm
3 4 ðIIDG Þ
1 , which re- þ þ
for Ne irradiation and 2.9 Å for He irradiation. If the
lates the grain size La in disordered graphite to the ratio of proportionality is attributed to charged impurity scattering
the integrated D and G band intensities ID and IG , and  [Eq. (1)], it would require addition of charge Z  4e per
the excitation wavelength (633 nm) [24]. Applying this incident ion. Figure 3(b) shows the density-independent
formula to our irradiated graphene gives La  60 nm, resistivity s for the same four experimental runs; s is
larger than the expected defect spacing of 10 nm, but very small (on order 10
3 h=e2 ) and does not change sig-
comparable to the transport mean free path of 50 nm nificantly with ion irradiation dose.
(see below). Figure 4(a) shows the change in the voltage of the
Figure 2 shows the ðVg Þ curves measured for the minimum conductivity Vg;min as a function of the inverse
pristine sample and following sequential Neþ irradiation mobility 1= (proportional to ion dose) for the four ion
doses at T ¼ 41 K in UHV, which is one of the four irradiation runs. For comparison, the magnitude of Vg;min
experimental runs shown in this Letter. Also shown are for potassium (K) dosing (addition of charged impurities)
predictions from Eq. (2) with the experimentally extracted is also shown (data from Ref. [19]), which is 5 times larger
defect radius R at nd ¼ 7:22  1011 cm
2 (see below). than a similar concentration of ion irradiation. Note that
Mobility  and the minimum conductivity min partially Vg;min is positive for ion irradiation, and negative for K
recover after heating to 485 K between each run, possibly dosing. Figure 4(b) shows min vs  for the same four ion
due to annealing or passivation of the defects. To determine irradiation runs and the K dosing run [19]. In sharp contrast
, and the resistivity s due to weak point disorder, the to the charged impurities introduced by K dosing, where
ðVg Þ curves are fitted to the form ðVg Þ
1 ¼ ½cg ðVg
min ¼ n ec varies slowly and nonmonotonically be-
Vg;min Þ
1 þ s [18]. We fit the hole side of the ðVg Þ cause n increases with increasing dose (decreasing ),
curve (Vg < Vg;min ) because the data span a wider Vg ion irradiation has a large effect on min , reducing min
range. Figure 3(a) shows 1= vs ion dosage for four roughly proportional to .
experimental runs on two different graphene samples as We now discuss the changes in ðnÞ upon ion irradia-
well as behavior for charged impurities [19,20]. For the tion. The density-independent resistivity [Fig. 3(b)] s 
irradiated samples, 1= increases linearly with ion dosage 3  10
3 h=e2 and is roughly independent of ion dose;
as expected for uncorrelated scattering. Fitting yields a at a carrier density of 1012 cm
2 , this corresponds to a
proportionality of 7:9  10
16 V s for the Neþ irradiation mean free path >2 m. The dominant signature, linear
236805-2
 week ending
PRL 102, 236805 (2009) PHYSICAL REVIEW LETTERS 12 JUNE 2009

Sample 1
st +
1 Ne irradiation (41K)
nd +
2 Ne irradiation (78K)
10 Sample 2
st +
1 He irradiation (10K)
nd +

|Vg,min| [V]
2 He irradiation (10K)
Sample 3
K dosing

(a)
1 10
2
1/µ [V s/m ]

σmin [e /h]
2
2
FIG. 3 (color online). (a) Inverse of mobility (1= ) vs ion
dosage for two Neþ irradiation runs on sample 1 and two Heþ (b)
irradiation runs on sample 2. The dashed line represents behavior 0
0.0 0.2 0.4 0.6 0.8
for the same concentration of charged impurities (potassium on 2
graphene from Ref. [19]). (b) Density-independent resistivity s µ [m /V s]
vs ion dosage.
FIG. 4 (color online). (a) Magnitude of the shift in the gate
voltage of minimum conductivity (jVg;min j) vs inverse mobility
ðnÞ ¼ ned with d independent of n, indicates that ion (1=). The shift is with respect to the initial value of Vg;min ,
irradiation either creates midgap states or charged impuri- 8.8 V and 6.4 V for the Neþ and Heþ irradiated samples,
ties. However, several observations argue that the observed respectively. (b) Minimum conductivity (min ) vs  for two
changes in ðnÞ are dominated by midgap states: (1) The Neþ irradiation runs on sample 1 and two Heþ irradiation runs
intervalley scattering observed in Raman spectroscopy on sample 2. Data for potassium dosing (Ref. [19]) are shown for
(Fig. 1) with scattering length on order 60 nm is incon- comparison. Vg;min is positive for ion irradiation, negative for K
sistent with s , but consistent with the associated mobility dosing.
 ¼ 1300 cm2 V
1 s
1 , at an ambient doping level of
1013 cm
2 , from which we calculate a mean free path
l  50 nm. This correspondence suggests that the trans- cies generated by ion knockoff [23]. Using this value of R
port mean free path significantly probes intervalley scat- in Eq. (2) yields a ðVg Þ similar in magnitude to the
tering from lattice defects. (2) The sign Vg;min for ion experimental curve, but with a stronger sublinearity
irradiation is positive, opposite to the expectation for dep- (Fig. 2). We do not understand this discrepancy, but it
osition of positive ions near the graphene and also opposite may be related to carrier density inhomogeneity persisting
to what was observed for ion-irradiated metal-oxide- to carrier densities much larger than n [27], or to the
semiconductor field-effect transistors [25]. (3) The reduc- addition of a small amount of deep charged impurities
tion in mobility, if due to charged impurities, would require [25] which would contribute a supralinear ðVg Þ. As dis-
4 added charges per incident ion, while Vg;min indicates cussed in the EPAPS material [21], the possible trapped
only 1=5 of a net charge per incident ion; this would noble-gas atoms are not likely to contribute significantly to
require a delicate balance between creation of positive and the resistivity in our irradiated graphene samples.
negative impurities, and such balance would need to hold Last we discuss the possibility of a metal-insulator
for incident Neþ and Heþ , which have very different transition in graphene with defects. Disorder-free graphene
momenta. (4) Within the Boltzmann transport picture, is expected to have a minimum conductivity of 4e2 =
h [7].
min ¼ n e [20] where the total mobility  ¼ ðd
1 þ The introduction of intravalley scattering only (e.g.,
c
1 Þ
1 . The roughly proportional relationship between charged impurities) is expected to induce weak antilocal-
min and  for ion-irradiated samples indicates that n , ization, increasing the conductivity [7,8] with decreasing
which is a function of nc , is nearly independent of ion temperature. However, intervalley scattering (which gives
dose [26]. rise to the Raman D band) is expected to induce weak
We therefore conclude that the data of Fig. 3(a) are localization, and insulating behavior, i.e.,  ! 0 as T ! 0,
dominated by uncharged lattice defects in graphene. The in graphene [7,8]. From Fig. 4(a), we can see that min in


impurity radius R  2:3 A–2:9
obtained from the linear
A ion-irradiated samples can be reduced well below 4e2 =
h,
fits of Fig. 3 is a reasonable value for single-carbon vacan- the minimum metallic value. Figure 5 shows the conduc-
236805-3
 week ending
PRL 102, 236805 (2009) PHYSICAL REVIEW LETTERS 12 JUNE 2009

100 (C. J.). The MRSEC SEFs were used in this work.
Irradiated Sample 1
Vg-Vg,min = Infrastructure support has also been provided by the
0V UMD NanoCenter and CNAM.
10V
30V

Pristine Sample
σ [e /h]

Vg-Vg,min = [1] C. Lee et al., Science 321, 385 (2008).

2

10 0V [2] K. S. Novoselov et al., Proc. Natl. Acad. Sci. U.S.A. 102,

10V 10 451 (2005).
30V [3] M. Ishigami et al., Nano Lett. 7, 1643 (2007).
[4] E. Stolyarova et al., Proc. Natl. Acad. Sci. U.S.A. 104,
9209 (2007).
[5] T. Stauber, N. M. R. Peres, and F. Guinea, Phys. Rev. B 76,
205423 (2007).
[6] M. Hentschel and F. Guinea, Phys. Rev. B 76, 115407
(2007).
10 100 [7] J. H. Bardarson et al., Phys. Rev. Lett. 99, 106801 (2007).
Temperature [K] [8] H. Suzuura and T. Ando, Phys. Rev. Lett. 89, 266603
(2002).
FIG. 5 (color online). Temperature dependence of the conduc- [9] S. Stankovich et al., Carbon 45, 1558 (2007).
tivity ðTÞ of pristine (open symbols) and irradiated (solid [10] S. Gilje et al., Nano Lett. 7, 3394 (2007).
symbols) graphene at three different gate voltages. ðTÞ taken [11] M. Eizenberg and J. M. Blakely, Surf. Sci. 82, 228 (1979).
on cooling is shown for sample 1 after run 1 (irradiation by Neþ , [12] A. Reina et al., Nano Lett. 9, 30 (2009).
dose 7  1011 cm
2 ) and annealing to T ¼ 300 K. ðTÞ for the [13] C. Berger et al., Science 312, 1191 (2006).
pristine sample is from Ref. [28]. [14] C. Thomsen and S. Reich, Phys. Rev. Lett. 85, 5214
(2000).
[15] A. C. Ferrari et al., Phys. Rev. Lett. 97, 187401 (2006).
tivity of the Neþ irradiated graphene sample as a function [16] R. Narula and S. Reich, Phys. Rev. B 78, 165422 (2008).
of temperature for three different gate voltages. The [17] N. H. Shon and T. Ando, J. Phys. Soc. Jpn. 67, 2421
T-dependent conductivity of pristine graphene from (1998).
Ref. [28] is also shown for comparison. The pristine gra- [18] C. Jang et al., Phys. Rev. Lett. 101, 146805 (2008).
phene has metallic behavior, e.g., d=dT < 0. However, [19] J.-H. Chen et al., Nature Phys. 4, 377 (2008).
even a small amount of irradiation (that changes the room- [20] S. Adam et al., Proc. Natl. Acad. Sci. U.S.A. 104, 18 392
temperature mobility <4  ) drastically affects the low- (2007).
temperature behavior. In stark contrast to graphene without [21] See EPAPS Document No. E-PRLTAO-103-003952 for a
irradiation, where min is largely temperature independent pdf format file which contains the description of experi-
mental method, the derivation of Eq. (2) in the main text,
from T ¼ 4–100 K [29], our irradiated sample is insulat-
the interpretation of the offset in the linear dependence of
ing with diverging resistivity as T ! 0. More work is
the inverse mobility on ion dosage, and the discussion on
needed to understand the exact nature of the insulating the possible trapped noble-gas atoms at the graphene-SiO2
state in ion-irradiated graphene, but the data are consistent interface. For more information on EPAPS, see http://
with the expectation that intervalley scattering produces www.aip.org/pubservs/epaps.html.
localization [7]. [22] G. M. Shedd and P. E. Russell, J. Vac. Sci. Technol. A 9,
In conclusion, we have measured charge transport in 1261 (1991).
graphene with defects induced by ion irradiation in ultra- [23] J. R. Hahn et al., Phys. Rev. B 53, R1725 (1996).
high vacuum. Defects cause significant intervalley scatter- [24] L. G. Cancado et al., Appl. Phys. Lett. 88, 163 106 (2006).
ing, as seen in a prominent Raman D band. Defects give [25] D. V. McCaughan and V. T. Murphy, J. Appl. Phys. 44,
rise to a constant mobility, with a magnitude 4 lower 2008 (1973).
than for similar concentration of potassium ions on gra- [26] Deep charged impurities (d  10 nm) would have a small
effect on n , while still changing , but should also result
phene, and consistent with scattering by midgap states. In
in a significantly superlinear ðnÞ for n > d
2 
contrast to charge impurity disorder, lattice defects reduce
1012 cm
2 [20], which is not observed. Experimentally,
the minimum conductivity dramatically, and produce an ion-induced charges in SiO2 typically lie within 5 nm of
insulating temperature dependence of the conductivity. the surface [25].
This work has been supported by NSF Grant No. DMR [27] E. Rossi and S. D. Sarma, Phys. Rev. Lett. 101, 166803
08-04976 (J. H. C., M. S. F., and E. D. W.), NSF-UMD- (2008).
MRSEC Grant No. DMR 05-20471 (J. H. C. and [28] J.-H. Chen et al., Nature Nanotech. 3, 206 (2008).
W. G. C.), and U.S. ONR Grant No. N000140610882 [29] K. S. Novoselov et al., Nature (London) 438, 197 (2005).

236805-4